A simple RF signal consists of a narrowband baseband signal which is modulated onto a high-frequency RF carrier. A more complex, broader bandwidth RF signal may consist of multiple narrowband signals modulated on similar RF carriers. A conventional RF digital data receiver uses an analog mixer and local oscillator (LO) to translate the signal from the carrier to a lower frequency, the baseband or intermediate frequency (IF), where it is then typically digitized and further signal extraction is performed in the digital domain. Similarly a conventional RF digital data transmitter works in the reverse direction, converting a digital baseband or IF signal to analog, followed by mixing it with the LO signal to upconvert the signal. It is then amplified for transmission through the antenna.
In many wireless communications systems, the antennas are mounted on a tall tower, and the digital processing is carried out in basestations located remote from the antenna, e.g., on the ground. It is not practical or efficient to carry GHz-range (microwave) RF signals long distances over coaxial lines, since signal attenuation is too high. Therefore, the conventional partitioning of the system places amplifiers and analog mixers on the tower with the antenna, as shown in FIG. 1. The lower-frequency baseband or IF signal can then be carried to and from the ground station with minimal attenuation loss using known and standard coaxial cable. For longer distances or broader bands, low-attenuation optical fibers have been used to carry this baseband or IF signal, typically in digital form.
Optic fiber systems capable of communicating microwave signals are known. These are typically applied in military radar applications.
A critical requirement in wireless communication systems is maintaining linearity on both transmission and reception. That is, the analog signal should not be subject to distortions that result in the alteration of signal composition within the band of interest. Spurious intermodulations from signal components within the band of interest can limit the useful dynamic range (called the spurious-free dynamic range or SFDR) of a receiver or transmitter system. Thus, in a narrow band system, non-linearities which result in spurious energy in frequencies outside of the band of interest are generally tolerable, since these are readily filtered or ignored. On the other hand, as the bandwidth of the signal of interest grows wider, the possibility of significant spurious signals in the band of interest due to non-linearities in the signal processing and transmission chain grows. These are not readily filtered or ignored, and therefore present limitations of the system.
In a radio system, non-linearities in the signal processing chain generally produce spurs or intermodulation products between components of the signal. The wider than the band of the signal processed, the greater the probability for intermodulation components (spurs) which lie within the band itself, and therefore cannot be eliminated with a band-pass filter. One strategy to avoid misinterpretation of such spurs is to employ a deconvolution process in the receiver to predict the effect of the non-linearity on the signal, and reverse its effects. However, this requires a receiver for the full bandwidth of the transmitter signal processing chain, which may be untenable. In general, it is preferred to avoid non-linearities, and where present, limit their effects as much as possible, so that the intermodulation products may be treated as noise.
A number of metrics specifying system performance are available. For example, the “Spur-free Dynamic Range” or SFDR, is a commonly used metric which presumes a relatively small number of high power signal components which produce intermodulation distortions, and therefore the parameter specifies the dynamic range of the system before spurs interfere with signal interpretation.
Spurious Free Dynamic Range is the usable dynamic range of a system before spurious noise interferes or distorts the fundamental signal. SFDR is the measure of the difference in amplitude between the fundamental signals and the largest harmonically or non-harmonically related spur within the band of operation. A spur is any frequency bin on a spectrum analyzer, or from a Fourier transform. Spur-free dynamic range (SFDR), as generally used, attempts to define receiver dynamic range in terms of two undesired interferers and the receiver noise floor. The spur-free dynamic range is the difference in dB between the receiver noise floor and the level of each of equal-amplitude signals that produce an in-band spurious product equal in power to the noise floor. Generally, the receiver third-order intercept point is used to predict the spurious product, but often the second-order distortion dominates.
The SFDR specification overlooks several important factors which influence dynamic range. First, it attempts to model interference by using just two (or perhaps up to 4) interfering signals. This overcomes some of the objections to single-tone testing, but the real signal environment is usually populated by a multitude of signals. Second, it does not reveal the effects of reciprocal mixing or compression like the desensitization dynamic-range test. Third, it does not effectively test the effects of receiver input filtering (preselection). Finally, SFDR, as it is ordinarily specified, considers only the third-order distortion. In fact, for many receivers, especially those with modest input filters, the second-order products may dominate.
Another metric employed is the “Noise Power Ratio” or NPR, which models the signal as white noise within a band, and then measures the noise floor at narrow ranges within the band resulting principally from intermodulation distortion.
These metrics are therefore extremes of a continuum which seeks to characterize a system to determine the impact of intermodulation distortion on in-band signals; the SFDR measures a specific effect of a small number, e.g., 2-4 signals, while the NPR measures the statistical effect of essentially an infinite number of signals which appear as white noise.
In a wideband system, the true SFDR measurement becomes difficult, since it may be difficult to determine a worst case effect without testing each different combination of signals, and the presumptions typically made for narrow band systems regarding the dominance of the third order intermodulation product. On the other hand, white noise itself is a statistical process, and the time of measurement is a relevant factor. For example, over short periods, intermodulation components from various spectral components within the white noise may cancel or reinforce each other, leading to a misleading measurement. Since each of these metric is presented as a simple ratio (often specified in decibels), care must be exercised in interpretation.
As used herein, the SFDR is intended to encompass the ratio of one or more respective signal components of equal amplitude within a band and their largest in-band intermodulation product, of any order, but excluding artifacts, such as a superposition of multiple intermodulation products. In cases where the SFDR is specified for a composite system with band-limiting filters, the input signals may take any value within the permissible input range, while the measurements are made after band-limiting.
The performance of high power amplifiers with many carriers (>10) is normally tested using a noise power ratio (NPR) measurement technique. In this test, white noise is used to simulate the presence of many carriers of random amplitude and phase. In a traditional test setup, the white noise is first passed through a bandpass filter (BPF) to produce an approximately square spectral pedestal of noise of about the same bandwidth as the signals being simulated. This signal is then passed through a narrow band-reject filter to produce a deep notch (typ. >50 dB) at the center of the noise pedestal. This noise signal is used to excite the test amplifier, which produces intermodulation distortion products that tend to fill in the notch. The depth of the notch at the output of the amplifier can be observed with a spectrum analyzer, and is the measure of the NPR.
NPR can be considered a measure of multi-carrier intermodulation ratio (C/I). NPR differs from multi-carrier C/I in that it is the ratio of carrier plus intermodulation to intermodulation (C+I/I). At higher ratios (C/I>20 dB), the two measures will approach the same value. The bandwidth of the noise source should be much wider the bandwidth of the BPF to insure the statistical distribution of the noise power resembles a random phase multicarrier source. The width of the noise pedestal is usually made equal to bandwidth of the channel under test. The width of the notch should be about 1 percent or less of the width of the noise pedestal.
As used herein, the NPR is intended to encompass the dynamic range of a white noise signal representing the full band range of a permissible input signal, with uniform amplitude, with an output signal from which the input signal is subtracted. Traditionally, such a measurement is obtained by providing a notch filter at the input for a narrow frequency, and then measuring the output within that notch band. On the other hand, digital processing techniques permit a more rigorous characterization. For example, a notch filter is not necessary; the output across the band can be compared with the input, and deviations characterized. In that case, the deviation can be expressed as a ratio to the input signal, and for example, a worst case deviation reported as the noise power ratio. In some cases, especially across a large band, the input amplitude power is not uniform. Since signal from one communication stream may interfere with signal from another, which has a different power spectrum, the NPR may be functionally specified based on real signal types, which may differ from equal amplitude noise across the band.
Thus, for example, in a wideband system which handles multiple smaller bands concurrently, a theoretical NPR may not yield fully functionally valuable information as compared to signal models. Typical signal types of interest are cellular communications, which include CDMA (and variants), GSM, LTE, WiMax, etc. Therefore, in addition to characterizing the performance with respect to a white noise across the wide band, it may be useful to determine a bit error rate (BER) of a specified model signal at a specified attenuation, or alternately, the limiting attenuation until a specified BER is achieved, in the presence of other types of signals in the remainder of the band. Since this type of characterization is communication protocol type-specific, it would not be generally useful for generic systems. Note that in some cases, the relative amplitude of signals within the broadband signal may be adjusted to optimize net performance; for example, if one signal interferes with that in an adjacent band, and both are not required to operate at maximum power, then one may be attenuated with respect to the other to achieve acceptable performance for both.
A primary source of nonlinearity in conventional transmission systems is the power amplifier (PA). In general, highly linear PAs (including semiconductor transistor PAs) are energy-inefficient, large and heavy. PAs that are compact and efficient tend to be highly nonlinear. A modern wireless communication system that requires many antennas and PAs on a given transmission tower may use, or even require the use of, relatively nonlinear PAs.
This nonlinearity limits the performance of the transceiver system, and must be corrected. Several linearization techniques are known in the prior art. Most of these techniques operate on the baseband signal before upconversion (see FIG. 1), or equivalently on the amplitude and phase of the RF signal. One such technique is predistortion, wherein a signal is deliberately distorted in such a way seeking to cancel the distortion that would be generated by the PA. These linearization techniques are generally limited to narrow-band signals, and furthermore the nonlinearities tend to become worse as the bandwidth increases.
It is well known in the prior art that there are important potential advantages to combining multiple RF signals in nearby bands to create single broadband signal. For example, this could substantially reduce the number of RF components required. However, the problems associated with nonlinear PAs and the inability to sufficiently correct their distortions severely limit the bandwidth of such a combined signal to less than about 100 MHz. Conventional technology does not have a solution to this problem, which makes broader band RF systems impractical.
Recently, a superconducting electronic technology (known as rapid-single-flux-quantum logic or RSFQ) has been demonstrated that can provide direct digitization of RF signals, as well as ultrafast digital processing at clock rates up to 40 GHz, with rates up to 100 GHz expected in the near future. This permits representation of RF signals in a directly digitized format, at broadcast frequencies, referred to as “Digital RF”™ (Hypres Inc.), wherein the sample rate is much larger than the carrier frequency, e.g., above the Nyquist rate. This provides a promising approach to very broadband multi-carrier RF communication signals (See, U.S. Pat. Nos. 7,280,623, 7,365,663, 7,362,125, 7,443,719, expressly incorporated herein by reference). However, RSFQ is an ultra-low-power technology (sub-mV level signals) that requires cooling to temperatures of 4 K (−269 C). Existing refrigeration technology makes it impractical to place these cryocooler systems on cellular towers, so they must be placed in ground stations. In that case, transmission losses of the RF signals between the ground and the tower largely eliminate the potential performance advantages. On the other hand, if the existing cryocoolers were sought to be placed on the tower, this would increase wind loading and power consumption on the tower, and potentially lead to increased maintenance costs due to potential servicing of the cryocooler in an inaccessible location. Since the tower electronics are standardized for various applications, even in cases where the tower components are accessible or increased maintenance and service permissible, the lack of economies of scale still make such installations infeasible.
Optical fibers would seem to be a promising alternative, since attenuation losses on fibers are quite small. However, the electro-optic (E/O) and opto-electronic (O/E) converters for prior-art multi-GHz analog signals had severe limitations in performance, again greatly reducing the performance advantages of the superconducting system. These converter limitations reflect both limits in bandwidth and limits in linearity. It is difficult to simultaneously achieve high bandwidth and high linearity, particularly in a reliable and inexpensive manufactured product. See, for example, U.S. Pat. No. 7,424,228, expressly incorporated herein by reference. There are several known methods for reducing nonlinearity (see, for example, U.S. Pat. No. 7,426,350, expressly incorporated herein by reference), but this remains a difficult problem.
The distinctions between digital and analog E/O (and O/E) modulators should be appreciated, which are directly analogous to the distinctions between digital and analog amplifiers. Digital amplifiers and modulators are nonlinear one-bit threshold devices, where only bit-errors are important. In contrast, analog amplifiers and modulators are generally required to be linear over a wide dynamic range, while nonlinearities generate intermodulations that reduce the useful spur-free dynamic range (SFDR). Digital optical modulators are long established for high-speed digital optical communications. In contrast, the technology of linear optical systems at microwave frequencies (microwave photonics or radio-over-fiber, or “RoF”) has been developing only over the past decade, with further improvements in performance continuing. For E/O converters, direct modulation of diode lasers is typically used for RF frequencies below a few GHz, while electro-optic interferometric modulation is used for higher frequencies. Photodiodes are used for O/E converters. The nonlinearity associated with such an optical link increases sharply with increasing bandwidth. A typical optical link specification may be expressed by SFDR=100 dB/B2/3, corresponding to 3rd-order intermodulation, where B is the bandwidth in dB-Hz. For a 10 GHz signal, corresponding to 100 dB-Hz, the link SFDR would be reduced to 33 dB. This would likely be quite sufficient for a digital signal, but would be too low for a high-dynamic range analog signal.
The low-power aspect of this superconducting technology requires the use of high-gain, high-bandwidth semiconductor amplifiers, which are near the limits of semiconductor technology, and commercial product is generally unavailable. Various non-ideal aspects of these amplifiers, especially nonlinear distortion, tend to degrade the performance of the overall system. Very recently, the same superconducting RSFQ technology has demonstrated the capability to provide very broadband digital predistortion directly on the digitized RF signal, for a digital-RF transmission system with a nonlinear PA. See, e.g., U.S. Pat. No. 7,313,199, expressly incorporated herein by reference. This can achieve much greater improvement in broadband system SFDR than would be possible using conventional baseband predistorters, and might be used to make these broadband systems practical. However, the link between the ground and tower units remains a problem.